The recent outbreak of the Wuhan virus inspired us to explore the models used for infectious disease spreading.
What it does
The model is based on a traditional model named Susceptible-Infectious Model, and we extend it further using the Markov model to allow for the dynamic, cross-boundary inference. Furthermore, we combined the Markov model with machine-learning algorithms so that the model's most important coefficient is learned from past timestamps and other features, therefore enhancing the performance of the traditional model.
How we built it
We used Python for the data science part of the project and used React to build a frontend, where we can visualize the prediction result of our data model.
Challenges we ran into
Difficulty to obtain and clean various data sources, deduce a novel machine learning model and construct the project from scratch.
Accomplishments that we're proud of
We devised a novel model that extends the Susceptible-Infectious model using Markov methods and introduced machine learning algorithms to achieve acceptable prediction results against the 2003 SARS disease outbreak.
What we learned
How to design and implement a machine learning model from scratch. How to combine the domain-specific model with modern machine learning technologies to achieve better results.
What's next for datathon-2020
Access the model's performance using the recent Wuhan disease outbreak. Apply the methodology to other infectious disease models to achieve better prediction results.